Configurations of resourceful and demanding attributes of organizational culture in US hotels: An innovative approach using topic modeling and fsQCA

Lee, Kyoung-Joo; Choi, Sun-Yong Article Configurations of resourceful and demanding attributes of organizational culture in US hotels: An innovative approach using topic modeling and fsQCA Journal of Innovation & Knowledge (JIK) Provided in Cooperation with: Elsevier Suggested Citation: Lee, Kyoung-Joo; Choi, Sun-Yong (2024) : Configurations of resourceful and demanding attributes of organizational culture in US hotels: An innovative approach using topic modeling and fsQCA, Journal of Innovation & Knowledge (JIK), ISSN 2444-569X, Elsevier, Amsterdam, Vol. 9, Iss. 4, pp. 1-13, https://doi.org/10.1016/j.jik.2024.100582 This Version is available at: https://hdl.handle.net/10419/327484 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ Configurations of resourceful and demanding attributes of organizational culture in US hotels: An innovative approach using topic modeling and fsQCA Kyoung-Joo Lee a , Sun-Yong Choi b, * a Department of Tourism Management, Gachon University, 1342 Seongnamdaero, Sujeong-gu, Seongnam, Gyeonggi-do, Republic of Korea b Department of Finance and Big Data, Gachon University, 1342 Seongnamdaero, Sujeong-gu, Seongnam, Gyeonggi-do, Republic of Korea ARTICLE INFO Article History: Received 3 April 2023 Accepted 24 September 2024 Available online 3 October 2024 ABSTRACT This study distinguishes itself from existing organizational culture studies by investigating the relatively under-studied theoretical relationship between organizational culture and employer attractiveness. To capture the holistic nature of organizational culture theory, this study adopts configurational analysis, which contrasts resourceful cultural attributes (collaborative, employee development, and fair-compensation cultures) with demanding attributes (result-oriented, overworked, and job-insecurity cultures). This study proposes three configurational propositions of employer attractiveness and employs review-based topic modeling and fuzzy-set quality comparative analysis (fsQCA) to overcome the limitations of traditional survey-based measurement and regression analysis. For the empirical analysis, this study constructs an industry-wide dataset comprising 2209 quarterly samples from 157 hotels over six years, utilizing 54,889 employee reviews posted on Glassdoor in the United States. Topic modeling analysis adopting Latent Dirichlet Allocation extracts the probabilities of six cultural attributes. Finally, the fsQCA generates three groups of 13 configurations, leading to employer attractiveness: a fully resourced culture, a resourced and lowdemanding culture, and a fairly compensated overwork culture. The findings confirm the core concepts of configurational analysis in contrast to regression analysis and present novel theoretical, methodological, and practical implications. © 2024 The Authors. Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Keywords: Organizational culture Configuration Employer attractiveness Latent Dirichlet Allocation Fuzzy-set quality comparative analysis JEL classification: C55 C63 C81 D23 J28 Introduction The success of organizations is heavily dependent not only on product and service market performance to win customers but also on the competitiveness of the labor market to attract competent employees (Gehrels & de Looij, 2011;Theurer et al., 2018). Thus, enhancing employer attractiveness and improving employer branding are essential strategic approaches to enhance the loyalty of current employees and attract prospective employees (Gehrels & de Looij, 2011;Lee & Choi, 2022;Srivastava & Bhatnagar, 2010;Theurer et al., 2018). Employer attractiveness refers to the benefits and values that both current and prospective employees perceive in a specific organization (Gehrels & de Looij, 2011;Lee & Choi, 2022;Theurer et al., 2018). This study focuses on organizational culture as a critical predictor of employer attractiveness. Organizational culture studies have long investigated the impact of cultural attributes on internal organizational members to explain organizational effectiveness, performance, and competitiveness (Hober et al., 2021;Jiang et al., 2023;Naveed et al., 2022;Quinn & Rohrbaugh, 1983;Schein, 1990;Schulte et al., 2009; Velasco Vizcaíno et al., 2021). By exploring a relatively unearthed theoretical relationship, we propose that organizational culture can play a critical role in developing and maintaining an organization’sattractiveness as an employer in the external labor market (Catanzaro et al., 2010;Lee & Choi, 2022;Sommer et al., 2017). The hotel industry faces tremendous challenges owing to a shortage of skilled workers and a high turnover rate. Therefore, acquiring talent and retaining fully motivated employees are important success factors for hotels (Gehrels & de Looij, 2011;Lee & Choi, 2022). According to a survey by the Bureau of Labor Statistics in the United States (US), the turnover rate in the hospitality sector increased significantly from 66.7 % in 2014 to 78.9 % in 2019, although the national average turnover rate in 2019 was 36.4 % (Lee & Choi, 2022). A problematic aspect of labor market competition for hotel firms is that it is arduous to differentiate themselves from competitors as attractive employers because jobs within the same hotel industry are largely * Corresponding author. E-mail addresses: [email protected] (K.-J. Lee), [email protected] (S.-Y. Choi). https://doi.org/10.1016/j.jik.2024.100582 2444-569X/© 2024 The Authors. Published by Elsevier España, S.L.U. on behalf of Journal of Innovation & Knowledge. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/) Journal of Innovation & Knowledge 9 (2024) 100582 Journal of Innovation &Knowledge https://www.journals.elsevier.com/journal-of-innovation-and-knowledge similar (Lee & Choi, 2022;Lievens & Highhouse, 2003). Given these issues, understanding hotels’attractiveness as employers is critical for building the most qualified applicant pool. To elucidate hotels’employer attractiveness, we investigated organizational culture’s role in determining recruitment performance and applicants’job acceptance. Beyond existing perspectives that highlight the impact of organizational culture on organizational effectiveness and performance, our study adds to the theoretical position that an attractive organizational culture can appeal to talented employees, encourage applicants to accept a job, and positively influence their tenure once hired (Catanzaro et al., 2010;Lee & Choi, 2022;Sommer et al., 2017). Furthermore, to investigate organizational culture, we adopted a configurational analysis of culture to materialize the theoretical potential of the holistic characteristics of organizational culture. The configurational analysis in this study intensifies our understanding of the complex combinational impacts of diverging cultural attributes, overcoming the limitations of a variable-centered approach using regression analysis of specific cultural attributes (Jung et al., 2009;Marinova et al., 2019;Misangyi et al., 2017;Ostroff & Schulte, 2014;Schulte et al., 2009). This study assumes complex combinations of resourceful (collaborative, employee development, and fair-compensation culture) and demanding (result-oriented, overwork, and job-insecurity culture) attributes of organizational culture and presents three configurational propositions for hotel employer attractiveness (Bakker & Demerouti, 2007;Lee & Choi, 2023,2024). In addition to the distinctive theoretical perspectives, we adopted a novel combination of methodological approaches. We performed review-based topic modeling and fuzzy-set quality comparative analysis (fsQCA) of organizational culture, overcoming the limitations of traditional survey-based measurement and regression analyses in the existing literature. In previous studies on organizational culture, a variablecentered approach performed a regression analysis of specificcultural attributes using self-reported employee surveys in a limited number of organizations (Lee & Choi, 2022;Marinova et al., 2019;Ostroff & Schulte, 2014). However, the existing methodological approach has numerous restrictions in fully materializing the theoretical potential of holistic organizational cultures. With its anthropological roots, the theory of organizational culture has adopted a holistic perspective from its early studies, suggesting that the combined patterns of cultural attributes cause different outcomes depending on how they are arranged (Fiss, 2007;Ostroff & Schulte, 2014). To address these limitations, we employed fsQCA for the configuration analysis of organizational culture. fsQCA, utilizing set-theoretic analysis, enables researchers to identify sufficient or necessary conditions to explain outcomes and determine the relative importance of different combinations of conditions (Chen et al., 2023;Fiss, 2007; Ostroff & Schulte, 2014;Pappas & Woodside, 2021). It provides a more fine-grained analysis for evaluating the importance of cultural dimensions in a configuration by distinguishing between core and peripheral dimensions (Fiss, 2007;Ostroff & Schulte, 2014;Pappas & Woodside, 2021). Furthermore, our study used an industry-wide employee review of hotels posted on Glassdoor, one of the largest job search platforms in the US (Corritore et al., 2020;Lee & Choi, 2022;Sull, Turconi, & Sull, 2020), which overcame the limitations of individual self-reporting surveys in a small number of organizations (Denison et al., 2014). We prepared a sample dataset consisting of 54,889 employee reviews from 157 hotel firms and aggregated personal reviews to evaluate corporate-level organizational culture with minimum personal bias. To measure the diverse dimensions of organizational culture in industry-wide big data, we adopted Latent Dirichlet Allocation (LDA), a topic modeling software (Blei, 2012;Corritore et al., 2020;Lee & Choi, 2022;Maier et al., 2018), which calculates the probability of topics for organizational culture. The theoretical relevance and validity of each topic were estimated using manual coding. Exploring the relatively unearthed theoretical relationship between organizational culture and employer attractiveness, this study draws on aconfigurational analysis to actualize the theoretical potential of the holistic nature of organizational culture. To achieve this goal, we adopted a novel research methodology of review-based topic modeling and fsQCA for the US hotel industry. The analysis results not only explore new theoretical potentials of organizational culture but also present a comparative analysis between fsQCA’sconfigurational analysis of holistic organizational culture and the regression analysis of individual attributes, generating novel methodological implications and practical insights for hotel employers. Literature review Organizational culture refers to the set of shared values, beliefs, norms, basic assumptions, and behaviors among organizational members that significantly influence how employees interact with each other and external stakeholders (Lee & Choi, 2022;2023,2024; Naveed et al., 2022;Quinn & Rohrbaugh, 1983;Schein, 1990;Schulte et al., 2009). Focusing on different influences of diverse cultural attributes, researchers have long investigated how organizational culture affects critical organizational outcomes, such as organizational effectiveness, performance, and competitiveness (Hober et al., 2021;Jiang et al., 2023;Lee & Choi, 2023,2024;Naveed et al., 2022; Quinn & Rohrbaugh, 1983;Schein, 1990;Schulte et al., 2009;Velasco Vizcaíno et al., 2021). Although existing literature primarily focuses on the cultural impact on internal members to explain critical organizational outcomes, it is crucial to recognize that distinctive cultural attributes can also send important signals to the external labor market, affecting an organization’s attractiveness as an employer (Catanzaro et al., 2010;Lee & Choi, 2022;Sommer et al., 2017). Recognizing this theoretical potential to extend to employer branding strategy, Sommer et al. (2017) conducted a scenario-based experiment in Germany. Their findings revealed that an innovative culture resonates more strongly with innovative individuals. Catanzaro et al. (2010) investigated gender differences in the context of competitive and supportive organizational cultures that affect job applicants. Based on the hypothetical organizations depicted in recruitment brochures in the US, most respondents were found to prefer working for a supportive organization. However, men were more inclined to pursue jobs in competitive cultures. Finally, Lee and Choi (2022) collected employee review data from hotel firms in the US, applied topic modeling analysis, and conducted a regression analysis. They reported that a set of positive organizational cultures, such as collaboration, employee development, fair-compensation, and customer focus, positively affect employer attractiveness. However, innovation culture has no significant effect. Despite the research efforts discussed above, studies on the impact of organizational culture on employer attractiveness remain theoretically and empirically limited. Organizational culture possesses holistic and systematic characteristics as one of the most significant aspects. Organizational culture combines different and sometimes conflicting attributes within an organization, and each individual attribute should be regarded as part of the entire culture (Lee & Choi, 2022;Marinova et al., 2019; Ostroff & Schulte, 2014). The holistic perspective emphasizes that each dimension of organizational culture should not be treated independently as they interact in a complex manner to generate essential outcomes (Ostroff & Schulte, 2014). Despite the significant theoretical potential of the holistic nature of organizational culture, regression analysis, which dominates existing studies, has numerous limitations in capturing causal complexity of organizational culture (Marinova et al., 2019;Misangyi et al., 2017;Ostroff & Schulte, 2014;Schulte et al., 2009). Given the holistic and systematic nature of organizational culture, the configurational approach used in our study can better fit the analysis of organizational culture than regression models, effectively intensifying our understanding of the relationship between the K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 2 complex configurations of organizational culture and differential organizational outcomes (Marinova et al., 2019;Misangyi et al., 2017;Ostroff & Schulte, 2014;Schulte et al., 2009). Configurational propositions of hotels’employer attractiveness Focusing on analyzing the complex combinations and interactions of different cultural attributes in generating outcomes, configurational analysis emphasizes the equifinality of varying configurations, the synergistic effects of different attributes, and the nonlinearity of causation (Fiss, 2007;Misangyi et al., 2017;Ostroff & Schulte, 2014). The equifinality concept suggests that different configurations can generate the same final outcomes from different initial conditions and paths (Fiss, 2007;Misangyi et al., 2017;Ostroff & Schulte, 2014). The equifinality of organizational culture presumes that organizations can develop varying configurations of organizational culture with different sets of internal attributes to obtain the same outcomes. Synergistic interactions presume complex interactions between different attributes, suggesting that causally relevant outcomes result from the intersections between different forces and events (Misangyi et al., 2017). The nonlinearity of causation presumes that the cultural attributes that are causally associated in a configuration are not related or are even inversely related in different configurations (Fiss, 2007;Misangyi et al., 2017;Ostroff & Schulte, 2014). Furthermore, this suggests that the absence of any attribute, as much as its presence, can lead to an important outcome when combined with other attributes (Misangyi et al., 2017). Organizational culture encompasses various cultural attributes simultaneously, although they often conflict with one another to meet complex environmental needs. Within a cultural configuration, different attributes interact in a complex manner to generate distinctive configurations that affect employer attractiveness. An effective configurational analysis of organizational culture begins by categorizing distinctive configurations and identifying different cultural attributes that are commonly relevant to the critical outcomes in question (Marinova et al., 2019;Ostroff & Schulte, 2014;Schulte et al., 2009). As many hotels suffer from persistent problems, such as shortage of a skilled workforce and high turnover rate, employer attractiveness is a critical success factor for winning against competitors (Lievens & Highhouse, 2003). Employer attractiveness refers to the envisioned benefits and values that current and prospective employees gain when they work for a specific organization (Lievens & Highhouse, 2003). Strong employer attractiveness enhances the organizational commitment of current employees and effectively appeals to prospective employees (Gehrels & de Looij, 2011;Theurer et al., 2018). As organizational culture affects the development of employees’dedication and commitment to an organization, varying configurations of organizational culture are strongly associated with the extent to which hotels are attractive to employees and applicants (Srivastava & Bhatnagar, 2010). Resourceful and demanding attribute of organizational culture Organizational culture theory posits that different and conflicting cultural attributes coexist within an organization and that their complex combinations give rise to critical organizational outcomes (Marinova et al., 2019;Ostroff & Schulte, 2014). Considering the diversity of organizational cultures and the causal complexity of their interactions, a taxonomy that classifies these attributes into categories related to employer attractiveness can significantly reduce the interpretative complexity of cultural interactions, thereby promoting a more intuitive understanding of their distinctive impacts. Furthermore, a classification that includes both positive and negative cultural attributes enables researchers to overcome potential bias from viewing only one side and adopt a more holistic and balanced perspective when analyzing the complex interactions of contrasting organizational cultures. Regarding the taxonomy of organizational culture, Lee and Choi (2023,2024) proposed a useful framework that distinguishes between the resourceful and demanding attributes of organizational culture based on the job demands-resources (JD-R) model. The JD-R model emphasizes the interactions between the resourceful and demanding aspects of the job to predict not only employees’wellbeing and high performance but also their job stress and burnout (Bakker & Demerouti, 2007;Jang et al., 2017;Lee & Choi, 2023,2024). Drawing on Lee and Choi’s (2023,2024) taxonomy, this study classifies cultural attributes into two categories: resourceful and demanding attributes of organizational culture. This study proposes that resourceful cultural attributes enhance employer attractiveness, whereas demanding attributes negatively affect it. Resourceful attributes such as collaborative, employee development, and fair-compensation culture are strongly associated with employee satisfaction, well-being, and high performance, significantly improving hotel employer attractiveness (Lee & Choi, 2022, 2023,2024). Collaborative culture in organizations highlights the sharing of a common vision and norms among employees. It considers employees as people (Tjosvold & Tsao, 1989), and promotes knowledge sharing and learning (Yang, 2007). In an employee development culture, employees are empowered to acquire new knowledge and skills to meet new job requirements (Hassan et al., 2006; Kuvaas & Dysvik, 2009). Fair-compensation culture in organizations upholds the norms of fairness, not only in the process of making decisions, but also in the results of distribution decisions, thus promoting employees’continuous efforts to improve performances (Choi & Chen, 2007;Edgar et al., 2014;Namasivayam et al., 2007). In contrast, demanding attributes such as result-oriented, overwork, and job-insecurity cultures create stressful and high-pressure working conditions, thereby negatively affecting employer attractiveness. By carefully monitoring employee efforts (Coget, 2010;Kohli et al., 1998) and creating strong inward pressure (Mitchell et al., 2019), the result-oriented culture stresses the achievement of end results through goal setting. The overwork culture requires employees to dedicate extra time to their work beyond the standard, agreed-upon hours, which may escalate various work risks (Ishaq et al., 2021;Mazzetti et al., 2016). Under the job-insecurity culture, employees perceive powerlessness to maintain jobs and have widespread concerns about job loss and career development risks (La stad et al., 2015; Nikolova et al., 2018). Configurational proposition The equifinality concept of configurational analysis (Fiss, 2007; Misangyi et al., 2017;Ostroff & Schulte, 2014) suggests that multiple cultural configurations combining both resourceful and demanding attributes may coexist to achieve the same organizational results of high employer attractiveness. As different sets of cultural attributes have synergetic interactions (Fiss, 2007;Misangyi et al., 2017;Ostroff & Schulte, 2014), different configurations of cultural attributes may lead to identical results in enhancing employer attractiveness. Thus, hotels may develop different cultural attribute configurations and achieve the same outcomes to ensure employer attractiveness. Proposition 1. The attractive configuration of hotels’organizational culture has multiple sets of resourceful and demanding attributes as different combinations of the cultural attributes can lead to employer attractiveness. The JD-R model suggests contrasting roles between resourceful and demanding attributes in employee commitment and organizational satisfaction (Bakker & Demerouti, 2007;Jang et al., 2017), directly affecting employer attractiveness (Jang et al., 2017). Resourceful attributes provide employees with social support, growth opportunities, and fair rewards to obtain high job satisfaction K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 3 and performance. However, demanding attributes lead to burnout and turnover intention by increasing performance pressure, work overload, and job stability concerns. Thus, given the opposing effects, attractive configurations of organizational culture are likely to develop more resourceful than demanding attributes. Proposition 2. The attractive configuration of hotels’organizational culture encompasses more resourceful attributes than demanding attributes as the resourceful attributes positively contribute to employer attractiveness, whereas the demanding attributes negatively contribute to the attractiveness. According to the competing value framework (Marinova et al., 2019;Quinn & Rohrbaugh, 1983), organizational culture configurations may encompass contrasting attributes that work together to generate meaningful outcomes. Furthermore, configurational analysis involves the synergistic effects of different attributes as well as the nonlinearity of causation (Fiss, 2007;Misangyi et al., 2017;Ostroff & Schulte, 2014). To be effective, organizations must be simultaneously concerned with contrasting values and norms. Furthermore, a growing body of paradox theory suggests that resolving paradox and working with contradictions are critical for organizational survival and high performance (Schad et al., 2016). To be competitive, organizations should achieve multiple goals while maintaining high congruence between conflicting cultural norms and values. Thus, an attractive configuration of hotels’organizational culture can have both resourceful and demanding attributes that generate employer attractiveness. Proposition 3. The attractive configurations of hotels’organizational culture encompass conflicting attributes simultaneously as the congruence between the resourceful attributes and demanding attributes can generate employer attractiveness. Methodology Distinguishing itself from existing survey-based measurements and regression analyses, this study proposes a novel methodological strategy that collects employee review data from Glassdoor and performs topic modeling to measure organizational culture in review texts (Fig. 1). Topic modeling encompasses three phases: tokenization of text, generation of 100 LDA topics, and manual coding of topics. Based on the measurement data, this study conducted both correlation and regression analyses and configuration analysis using fsQCA, leading to a comparative discussion of the analysis results. The regression analysis assumes that the relationships are symmetrical and linear (Pappas & Woodside, 2021). It focuses on assessing the unique contribution of a cultural attribute while maintaining the influence of other cultural dimensions (Pappas & Woodside, 2021). With a symmetrical assumption, regression analysis aims to identify determinants that explain high levels of the outcome variable, assuming that the exact opposite will result in low levels of the same outcome (Pappas & Woodside, 2021). Consequently, regression analysis encounters difficulties in accounting for potentially complex interactions among the different cultural attributes that operate as a system (Ostroff & Schulte, 2014;Pappas & Woodside, 2021). Considering the differences in methodologies, we initially performed a regression analysis and then employed a configurational analysis using fsQCA, providing a comparative analysis between the two methodological approaches. The fsQCA process in our study consists of three phases: calibration, necessity and sufficiency analysis, and testing of the configurational proposition. A comparative analysis of the findings provided a crucial understanding of the methodological differences and hidden insights revealed by the configurational analysis. Collecting employee review data To perform an empirical analysis, we tapped into the Glassdoor website in the US and collected reviews of hotel firms written by current and former employees. Glassdoor, one of the largest job search platforms in the US, has accumulated 55 million reviews of 900,000 organizations since 2008 and recorded 67 million unique monthly visitors in 2020. After authenticating identity, Glassdoor enables users to review anonymously and search for job post reviews to gain employment details (Corritore et al., 2020;Lee & Choi, 2022,2023, 2024). This study combines all texts (both pros and cons) to identify employees’diverse cultural attributes. Many existing studies of organizational culture rely on surveys of a few individuals in a firm, raising concerns of personal bias and limited validity (Denison et al., 2014;Lee & Choi, 2022;2023;2024). Reviews on the Glassdoor Fig. 1. Research analysis map. K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 4 platform enable employees to provide free text responses about a firm’s organizational culture and help researchers gain a more comprehensive picture of a firm’s organizational culture than surveys (Sull, Turconi, & Sull, 2020). To search for hotel firms on Glassdoor, we adopted a list of hotel brands and chains in the US from the Looking for Booking website and used not only the direct terms such as “hotel(s),”“motel(s),” “lodge(ing),”but also indirect terms such as “resort(s),”“vacation(s),” “casino(s),”“park(s),”and “hospitality”for the search process. Initially, 78, 563 employee reviews were collected from 189 hotel firms, with a minimum of 100 reviews published in November 2020. Considering that this study assessed organizational culture at the firmlevel, it should have a sufficient number of employee reviews to evaluate cultural attributes appropriately (Corritore et al., 2020;Lee & Choi, 2022;2023;2024). Therefore, this study sets five reviews per quarter as the minimum requirement for the analysis. Furthermore, it restricts reviews written between January 2014 and December 2019 to avoid unusual business disturbances caused by COVID-19. After filtering and formatting the data, we constructed an industrywide dataset encompassing 2209 quarterly samples from 157 hotels over six years, including 54,889 employee reviews. Measurement by topic modeling analysis To measure organizational culture from review texts, this study adopted the LDA topic modeling technology and assessed the probability distribution of cultural topics in quarterly reviews. As an unsupervised machine-learning tool, LDA topic modeling enables users to identify the hidden structure of documents by generating interpretable topic distributions in the documents (Blei, 2012;Corritore et al., 2020;Lee & Choi, 2022,2023,2024;Maier et al., 2018). In the topic modeling process, LDA assumes a “bag of words”that ignores word orders in textual documents. LDA is designed to have a mixed-membership model of grouped data, where each group presents multiple topics in different proportions. The topic modeling identifies distinctive topics across the corpus in such a way that it collects words that frequently co-occur within each employee review. Subsequently, it generates a topic matrix that presents a probabilistic mixture of topics in each review by calculating the percentages across all the topics (Corritore et al., 2020;Lee & Choi, 2022;Maier et al., 2018). To measure hotel firms’organizational culture, we integrated the quarterly Glassdoor reviews of a firm and performed an LDA analysis that generated the probability of 100 cultural topics by adopting an unsupervised machine-learning process. Specifically, we conducted a three-step analysis (Lee & Choi, 2022;Maier et al., 2018). First, we performed basic and standard procedures for data cleaning and preprocessing of unstructured text. The cleaning process conducted text tokenization, such as discarding punctuations and word capitalizations, eliminating stop words, removing highly frequent and infrequent words, and performing stemming and/or lemmatizing (Lee & Choi, 2022;Maier et al., 2018). Second, 100 LDA topics were generated. Choosing the number of LDA topics generated by machine-learning tools is a highly complicated task, because there is no standard procedure for calculating the appropriate number of topics (Lee & Choi, 2022;Maier et al., 2018). Users can freely set the number of topics in the LDA operation (e.g., 30, 50, 100, and 500). A dilemma that researchers may face is that generating a small number of broad topics creates the problem of obtaining general topics containing different themes, whereas generating many topics leads to overly specific and narrow topics (Lee & Choi, 2022;Maier et al., 2018). Rather than running the risk of missing critical cultural topics, we generated a relatively large number of 100 topics, which helped gain narrowly defined topics for manual identification of diverse cultural attributes. Third, to identify cultural attributes in the LDA topics, we performed a manual coding process to estimate the theoretical relevance and validity of each topic (Table 1). Although unsupervised machine learning of LDA can generate multiple topics, it cannot offer the meaning of the topics and, thus, necessitates human interpretation to estimate the cultural relevance of the topics. The meanings of topics are not deterministic and must be evaluated based on substantive theoretical constructs (Lee & Choi, 2022;Maier et al., 2018). Thus, we adopted the most straightforward approach in which researchers read the terms in the model results, interpreted the meaning of topics, and matched them to cultural attributes (Corritore et al., 2020;Lee & Choi, 2022;2023;2024;Maier et al., 2018). During the matching process, the three researchers independently read, interpreted, and selected topics relevant to the theoretical constructs of cultural attributes. After examining 100 topics with the 15 most frequent terms, one researcher proposed 35 topics, another 39 topics, and a third 44 topics. To select independently coded topics, we adopted the decision rule suggested by Corritore et al. (2020) and selected topics for which at least two researchers independently matched the cultural attributes. The selection process performed by the three researchers resulted in 39 topics. As Table 1 shows, the conflicting attributes of organizational culture include multiple topics and key terms. Regarding resourceful attributes, collaborative culture has 12 topics, employee development has seven topics, and fair-compensation has 10 topics. For demanding attributes, results-orientation, overwork, and job insecurity encompassed two, five, and three topics, respectively. Finally, we added the quarterly probabilities of topics to measure the six cultural attributes at the firm level and analyzed the attractive configurations of organizational culture. This study’s outcome is the attractiveness of hotels as employers, and it adopted two indicators available in Glassdoor reviews as measurements (Fig. 1). First, Glassdoor allows users to rate “general employer satisfaction”on a five-point scale, which we adopted as an employer attractiveness outcome. Second, the response to the question, if the user “recommends the firm to friends”was also used as an outcome. We assigned 1 for a recommendation, 0 for neutral or no answer, and 1 for no recommendation. For a firm-level evaluation of employer attractiveness, we used the means of all quarterly ratings provided by the reviewers. Analysis results Correlation and multiple regression analysis To test the propositions, we adopted a two-step approach, performing a traditional correlation and regression analysis before configurational analysis using fsQCA. Table 2 presents the descriptive statistics and correlations. As predicted in the theoretical discussion, resourceful cultural attributes, such as collaboration, employee development, and fair-compensation culture, are positively and significantly correlated with employer attractiveness outcomes, such as employer satisfaction and recommendation. The result-oriented and job-insecurity cultures with demanding cultural attributes have a negative and significant correlation with employer attractiveness outcomes. However, contrary to theoretical prediction, overwork culture had no significant correlation with attractiveness outcomes, suggesting its complex relationship with other attributes in generating attractiveness outcomes. The regression model helps evaluate the influence of individual cultural attributes on employer attractiveness. As some cultural attributes can be more essential for determining the overall organizational culture and employer attractiveness outcomes, the regression model can help researchers assess the relative importance of individual attributes (Ostroff & Schulte, 2014). Furthermore, the preliminary regression analysis enables us to compare between the traditional variable-centered analysis and configurational approach, intensifying K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 5 our understanding about the strengths in configurational analysis of organizational culture. The regression analysis result in Table 3 demonstrates that all the resourceful attributes, such as collaborative (A: b= 0.263, p<.001; B: b= 0.200, p<.001), employee development (A: b= 0.217, p<.001; B: b= 0.224, p<.001), and fair-compensation (A: b= 0.180, p<.001; B: b= 0.158, p<.001), have positive and significant effect on both employer satisfaction and recommendation, respectively, thereby confirming the theoretical predictions. Additionally, the demanding attributes, such as result-oriented (A: b=.128, p<.001; B: Table 1 Construct measurement by topics generated by LDA. Construct Topic Frequent term Collaborative culture 3 friendly, environment, great, work, atmosphere, helpful, everyone, coworkers, team, worker, fun, sometimes, pleasant, relaxed, welcoming 5 work, love, great, amazing, people, really, intern, awesome, place, wish, enjoy, caring, hard, team, environment 8 make, everyone, help, feel, like, thing, way, every, willing, know, same, think, something, just, else 9 like, family, feel, stay, treat, part, friend, always just, well, number, come, definitely, enjoy, work 14 work, environment, place, great, really, positive, fun, team, enjoy, hard, professional, tough, exciting, supportive, chance 16 good, work, nice, people, far, coworker, bit, place, building, everything, load, area, definitely, heavy, generally 22 work, great, place, good, home, excellent, colleague, well, smart, think, social, clear, condition, tight, add 33 work, people, really, nice, get, hard, lot, time, super, cool, easy, fun, everyone, extremely, awesome 47 staff, friendly, management, short, always, work, sometimes, busy, good, shift, colleague, hour, environment, finish, fairmont 51 team, member, value, share, family, together, need, supportive, top, core, truly, well, part, wonderful, positive 53 environment, work, fun, great, good, fast, friendly, paced, challenge, supportive, colleague, overall. beautiful, professional, win 64 work, really, love, people, back, enjoy, face, help, everyone, laid, everyday, interact, able, community, appreciate Employee development culture 13 opportunity, room, advancement, growth, benefit, plenty, small, little, offer, compensation, pto, competitive, position, discounted, fulltime 23 company, opportunity, move, grow, growth, lot, willing, advance, city, career, offer, great, quickly, limit, limited 24 opportunity, career, growth, grow, development, great, culture, learn, challenge, company, excellent, progression, leadership, path, compensation 41 training, program, opportunity, development, excellent, property, great, growth, promotion, company, process, promote, career, white, engage 45 move, many, opportunity, position, department, different, transfer, around, location, lot, difficult, hard, advance, available, entry 61 experience, learn, lot, skill, different, gain, knowledge, difficult, great, interaction, training, exposure, variety, new, type 76 expectation, management, support, training, high, well, set, goal, expect, due, little, need, pressure, issue, unrealistic Fair-compensation culture 11 great, benefit, salary, travel, competitive, culture, amazing, package, company, rule, strict, awesome, minimal, corporate, advancement 18 hotel, discount, employee, room, rate, benefit, travel, perk, cafeteria, meal, advance, renovate, loyalty, paid, reduce 20 great, benefit, perk, awesome, location, atmosphere, sometimes, vary, strict, empower, diverse, staffed, throughout, starting, split 32 hour, good, pay, flexible, work, schedule, scheduling, flexibility, overtime, management, little, enough, manager, boss, inconsistent 34 company, great, benefit, reward, well, fantastic, amazing, offer, perk, excellent, recognize, vision, especially, recognition, hotels 42 good, benefit, pay, union, fair, salary, plus, employee, appreciation, provide, non, teamwork, community, lower, quality 44 good, work, salary, environment, professional, training, standard, experience, benefit, personal, facility, great, less, worldwide, learning 52 great, benefit, lot, good, atmosphere, environment, fun, beautiful, culture, upward, mobility, move, movement, potential, consistent 67 pay, job, good, people, easy, decent, enough, suck, get, nice, sometimes, low, slow, amount, work 96 benefit, health, insurance, 401k, discount, medical, plan, vacation, offer, dental, pay, match, include, travel, etc Result-oriented culture 80 sale, bonus, sell, make, base, product, incentive, performance, high, review, potential, market, money, goal, leave 87 money, make, cut, business, tip, staff, spend, need, server, save, restaurant, enough, cost, keep, budget Overwork culture 12 hour, long, work, great, pay, flexible, foot, demand, simple, unpredictable, hair, emotional, transport, excessive, log 17 shift, night, hour, late, off, work, early, time, schedule, free, holiday, day, request, weekend, cover 21 day, week, every, year, per, month, hour, two, off, get, work, even, three, few, raise 55 holiday, day, off, hour, weekend, season, time, long, schedule, expect, week, every, open, busy, four 79 pay, wage, low, minimum, hourly, hour, overtime, increase, less, tip, living, fair, barely, rate, uniform Job-insecurity culture 31 people, get, bad, just, know, ever, hire, leave, fire, right, keep, job, way, happen, look 35 job, work, difficult, security, keep, hard, responsibility, situation, able, deal, always, someone, help, find, come 71 high, turnover, rate, pay, employee, low, management, extremely, stress, position, pressure, co-worker, amount, decent, christmas Note:(Lee & Choi, 2022). Table 2 Descriptive statistics and correlations (N= 2209). Means S.D. 1 2 3 4 5 6 7 Collaborative .117 .029 Employee development .061 .021 0.056 ** Fair-compensation .078 .023 .003 .174 ** Result-oriented .017 .010 0.129 ** 0.052* 0.069 ** Overwork .052 .019 0.027 0.236 ** 0.078 ** 0.069 ** Job-insecurity .031 .014 0.142 ** 0.130 ** 0.131 ** .068 ** 0.050* Employer satisfaction 3.371 .584 .276 ** .242 ** .234 ** 0.192 ** 0.032 0.165 ** Employer recommendation .166 .348 .213 ** .249 ** .213 ** 0.177 ** 0.030 0.158 ** .829 ** Note. * and ** denote significance at the 5 % and 1 % levels, respectively. K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 6 b=.122, p<.001) and job-insecurity culture (A: b=.066, p<.05; B: b=.070, p<.05) have a negative and significant effect on employer satisfaction and recommendation, respectively. However, contrary to our theoretical prediction, an overwork culture has no negative impact on attractiveness outcomes. Configurational analysis by fsQCA For a configurational analysis of hotels’organizational culture, we adopted fsQCA using the fsQCA 3.0 software. The fsQCA approach examines the relationship between all possible combinations of predictors and the outcome of interest, generating various combinations of conditions that lead to the absence of outcomes, positive outcomes, or negative outcomes (Fiss, 2007;Pappas & Woodside, 2021; Ragin, 2009). Given our research goal of understanding the cultural configuration that generates employer attractiveness, we focused on identifying configurations leading to the presence of outcomes, specifically employer satisfaction and recommendation. First, we initially performed calibration, which was aimed at allocating the set memberships of each case under causal conditions, such as six cultural attributes and two employer attractiveness variables. In fsQCA, the calibration process assigns collective membership to cases using either direct or indirect methods. In direct calibration, researchers set exactly three qualitative breakpoints indicating fullset membership, full-set non-membership, and intermediate-set membership in the fuzzy set for each case. By contrast, in indirect calibration, the measurement is rescaled based on qualitative evaluations, and researchers may choose to calibrate a measure differently based on their investigation (Fiss, 2007;Ragin, 2009). The selection of either method depends on researchers’substantive knowledge of the data and the underlying theory (Fiss, 2007;Ragin, 2009). Considering that the procedure is clear for verification and replication by others, researchers have discretion in determining the procedure for assigning fuzzy values to cases and adopting threshold values. Because the data in this study do not completely follow a normal distribution and exhibit skewness, we set the full membership points of the six antecedents and two outcome variables as the upper quartile (75 %) of the case data, the intersection points as the median (50 %) of the case data, and the complete non-membership points as the lower quartile (10 %). Table 4 lists the anchor point calibrations. Moreover, in the fsQCA, cases with an intermediate set membership of exactly 0.5 are typically dropped. To address this issue, it is recommended to add a constant of 0.001 to the causal and outcome conditions below a full membership score of 1 (Fiss, 2007;Ragin, 2009). Thus, we added 0.001 to all predictor and outcome conditions after calibration was conducted. Second, based on the calibration, we performed both necessity and sufficiency analysis. On the one hand, before building a truth table, we examined whether any of the causal conditions could be considered necessary. The fsQCA tool enables researchers to detect necessary conditions that help facilitate the occurrence of an event. The consistency threshold of the necessity analysis is set at 0.9. Table 5 displays the results of the necessity analysis. No necessary consistency of the antecedent cultural conditions on employer satisfaction and recommendation is above 0.9, indicating that no single cultural attribute is necessary for attractiveness outcomes. In other words, a single cultural attribute has insufficient explanatory power for employer attractiveness. Therefore, this result led us to further analyze the combined effect of antecedent conditions. On the other hand, after confirming that no single cultural attribute constitutes a necessary condition for employer attractiveness, we analyzed the conditional combination of the six cultural dimensions to obtain employer attractiveness. To determine the appropriate causal combinations generating the outcome, the number of rows in the truth table should be reduced. We adopted two reduction criteria. First, considering the large sample size of our study, the minimum number of cases to consider a combination relevant was set at 12. Second, consistency criteria indicated the extent to which a combination of causal conditions is consistent in relation to the outcome. Table 3 Results of regression analysis (N= 2209). Model A Employer satisfaction Model B Employer recommendation Path coefficient T-statistic (P-value) Path coefficient T-statistic (P-value) Collaborative .263*** 13.478 (0.000) .200*** 10.042 (0.000) Employee development .217*** 10.755 (0.000) .224*** 10.891 (0.000) Fair-compensation .180*** 9.240 (0.000) .158*** 7.924 (0.000) Result-oriented 0.128*** 6.598 (0.000) 0.122*** 6.133 (0.000) Overwork .029 1.446 (0.148) .029 1.438 (0.150) Job-insecurity 0.066* 3.347 (0.001) 0.070* 3.480 (0.001) F-value 92.621*** 71.979*** R 2 .199 .162 Note. *, **, and *** denote significance at the 5 %, 1 % and 0.1 % levels, respectively. Table 4 Calibration percentile and statistics. Percentile .75 .5 .1 Employer satisfaction 3.7857 3.4286 2.5714 Employer recommendation .4 .2001 - 0.3333 Collaborative .1338 .1151 .1338 Employee development .0739 .06 .0739 Fair-compensation .0922 .0776 .0922 Result-oriented .0224 .0153 .0224 Overwork .0638 .0502 .0638 Job-insecurity .0395 .0303 .0395 Table 5 Necessity analysis. Employer satisfaction Employer recommendation Consistency Coverage Consistency Coverage Collaborative .6711 .6817 .6544 .6743 »Collaborative .4656 .5514 .4801 .5768 Employee development .6871 .6940 .6889 .7059 »Employee development .4605 .5489 .4572 .5529 Fair-compensation .6807 .6937 .6675 .6900 »Fair-compensation .4576 .5398 .4674 .5595 Result-orientation .5605 .5742 .5647 .5868 »Result-orientation .5719 .6707 .5677 .6754 Overwork .6031 .6142 .5969 .6167 »Overwork .5303 .6261 .5355 .6414 Job-insecurity .5735 .5785 .5767 .5901 »Job-insecurity .5597 .6682 .5548 .6719 K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 7 As the consistency threshold should not be less than 0.75 (Pappas & Woodside, 2021;Ragin, 2009), we set the consistency threshold at 0.89. After eliminating the combinations that do not meet the above two conditions from the truth table, Figs. 2 and 3present the results of fsQCA for employer satisfaction and recommendation, respectively. The output of the truth table typically presents three types of solutions: complex, intermediate, and parsimonious. To generate a valid truth table, researchers typically focus on both intermediate and parsimonious solutions and identify both core and peripheral conditions (Fiss, 2007;Pappas & Woodside, 2021;Ragin, 2009). A complex solution generates all possible combinations of attributes, making interpreting the solutions difficult. The parsimonious solution only presents the core conditions, which are essential conditions that cannot be excluded from any solution. The intermediate solution presents both the core and peripheral conditions by performing a counterfactual analysis of the complex and parsimonious solutions. Many researchers have presented final solutions by comparing and elaborating on parsimonious and intermediate solutions (Pappas & Woodside, 2021). However, in our analysis, all three solutions (complex, intermediate, and parsimonious) yielded identical solutions without distinguishing between the core and peripheral conditions. These identical solutions may result from the large dataset used in this study (Reichert et al., 2016). Previous fsQCA studies using large sample sizes have demonstrated this tendency. For example, Yang’s (2018) research with a sample of 2163 adults, Tho and Trang’s (2015) work with a sample of 843 students, and Reichert et al.’s (2016) analysis of 614 low-tech firms yielded identical solutions for all three types. Thus, we interpreted the solution results directly, without elaborating on possible scenarios, by comparing the core and peripheral conditions. Each column in Figs. 2 and 3demonstrates a combination of conditions or configuration of cultural attributes associated with employer attractiveness, such as employer satisfaction and recommendation. The black circle in the figure indicates the presence of a condition that leads to an outcome of interest; the crossed-out circle denotes the absence of a condition that causes an outcome; and the blank spaces represent a “don’t care”situation, suggesting that neither the presence nor absence of the condition is related to the outcome. At the bottom, the figures show the consistency, unique Fig. 2. fsQCA analysis for employer satisfaction (N= 2209). K.-J. Lee and S.-Y. Choi Journal of Innovation & Knowledge 9 (2024) 100582 8